Many Internet sellers now commit a very short lead-time for delivery upon the confirmation of orders. This requires to generate a delivery routing plan in real time. This thesis presents a solution framework that dynamically traces the buyers’ ordering process. The core is a genetic algorithm designed to consider multiple possible delivery scenarios at the same time. It generates a pool containing situations which are more likely to happen and adjust the pool according to current customers’ purchasing progress. With consideration of possible scenarios in advance, a delivery route with low cost can be obtained immediately after the cut-off time.

Keywords: Delivery with multiple scenarios, TSP, Genetic Algorithm